Research And Optimization For The Intra-prediction And Transform In H.264

With the development of digital image processing and computer network technology, the 3rd generation video surveillance system has come up with the symbol of source compressing, bus transport and image database etc. technologies. Its basic theory is using the video compressing technology immediately to compress the analog signals outputting by the video capture unit, then transport the encoded signals to the receiver by the computer network, and decode, play back, and store them.Video encoding technology is very important in the 3rd generation video surveillance system considering not only transporting, storing, but also image quality. A good encoding standard is needed especially when the network bandwidths for transport are taken to account.ITU-T H.264 / MPEG-4 Part 10 Advanced Video Coding (commonly referred as H.264/AVC) is the newest international video coding standards. It was developed by a Joint Video Team (JVT) consisting of experts from ITU-T's Video Coding Experts Group (VCEG) and ISO/IEC's Moving Picture Experts Group (MPEG). In the process, the standard was created that improved compression efficiency by a factor about two (on average) over MPEG-2. Several coding elements contribute to the great improvement in compression efficiency for the new standard. However, the computational complexity of the H.264/AVC encoder is dramatically increased by this optimization technique so that it is difficult to implement in real-time applications.The basic theories for video encoding and the development of video encoding standard are introduced first, and the main parts of H.264 are introduced in Chapter 2.The intra-prediction method is analyzed in detail in Chapter 3, and the computionally complex part - prediction mode selection is researched deeply, then a fast mode selection method based adaptive threshold is given. The threshold can be adjusted adaptively based on both the image texture and the QP value to determine whether the most probable mode can be selected as the optimal mode to end the selection process. To decrease the complexity of mode selection further, weuse a partionl comtuting method for the remaining modes. Some modes are tested first, and the neighbor modes of the temp-optimal mode are compared to get the best one. The simulation results show that without considerable performance degradation, the proposed method decreases complexity of the intra-prediction mode decision part in H.264, and saves the computing costs greatly. And the method supports SIMD instructions well, so it's very useful for assembly optimization in applications.The one-directional prediction in H.264 is analyzed further in Chapter 4, and we found that the one-directional prediction results in identical prediction values for all samples in the same direction. However in a nature video frame, one sample usually has different value from its neighbors, even in a smooth region. Furthermore, the value of a sample lies highly likely among the values of its surrounding neighbors, and thus bi-directional (forward and backward) prediction usually brings a better result than one-directional one. So the bi-directional prediction method is discussed and improved based on the idea proposed by Mr. Andy Teng from Canada in DCC'04.In Chapter 5, the structure characteristics and the software development flow of C6000 series DSPs are introduced after a brief introduction of TI TMS320C6000 DSPs, and then the assembly optimization of intra-prediction and transform is discussed in detail. The adaptive block-size transform of H.264 is analyzed and discussed in Chapter 6, and a brief summery and prospect is given in the last chapter.